Anmerkungen:
Diese Datenquelle enthält auch Bestandsnachweise, die nicht zu einem Volltext führen.
Beschreibung:
Proving timeliness is an integral part of the verification of safety-critical real-time systems. To this end, timing analysis computes upper bounds on the execution times of programs that execute on a given hardware platform. Modern hardware platforms commonly exhibit counter-intuitive timing behaviour: a locally slower execution can lead to a faster overall execution. Such behaviour challenges efficient timing analysis. In this work, we present and discuss a hardware design, the strictly in-order pipeline, that behaves monotonically w.r.t. the progress of a program's execution. Based on monotonicity, we prove the absence of the aforementioned counter-intuitive behaviour. At least since multi-core processors have emerged, timing analysis separates concerns by analysing different aspects of the system's timing behaviour individually. In this work, we validate the underlying assumption that a timing bound can be soundly composed from individual contributions. We show that even simple processors exhibit counter-intuitive behaviour - a locally slow execution can lead to an even slower overall execution - that impedes the soundness of the composition. We present the compositional base bound analysis that accounts for any such amplifying effects within its timing contribution. This enables a sound compositional analysis even for complex processors. Furthermore, we discuss hardware modifications that enable efficient compositional analyses. ; Echtzeitsysteme müssen unter allen Umständen beweisbar pünktlich arbeiten. Zum Beweis errechnet die Zeitanalyse obere Schranken der für die Ausführung von Programmen auf einer Hardware-Plattform benötigten Zeit. Moderne Hardware-Plattformen sind bekannt für unerwartetes Zeitverhalten bei dem eine lokale Verzögerung in einer global schnelleren Ausführung resultiert. Solches Zeitverhalten erschwert eine effiziente Analyse. Im Rahmen dieser Arbeit diskutieren wir das Design eines Prozessors mit eingeschränkter Fließbandverarbeitung (strictly in-order pipeline), der sich bzgl. des ...